The Structure of Robot-Directed Interaction compared to Adult- and Infant-Directed Interaction using a Model for Acoustic Packaging

摘要

Abstract — As it is often assumed that an interaction with a robot is similar to an interaction with an infant, one would expect similar characteristics in tutoring behavior in human-robot interaction from which the robot’s learning processes can benefit. Especially learning actions can profit from the use of language. As it has been shown by Hirsh-Pasek and Golinkoff introducing the idea of Acoustic Packaging, language not only serves as a social cue but also provides functional structure. In order to better understand how robot-directed interaction is structured and how the modifications may help to extract cues relevant for learning actions from a tutoring situation, in the present paper we analyzed the structure of Acoustic Packages in the adult-robot situation. Our results indicate that in human-robot interaction, the phenomenon of Acoustic Packaging is similar in adult-child interaction: In both situations, more Acoustic Packages with less content can be found, compared to adult-adult interaction. This can be interpreted as an encouraging result towards the goal of building robots that can learn from a tutoring situation. I.